# pentest-ai — Offensive Security MCP for Claude Code

> pentest-ai is a Python CLI and MCP server that lets Claude Code run verified probes, chain attack paths, and export reports for authorized testing.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

```bash
pip install ptai

# (optional) interactive wizard for MCP client config
ptai setup --mcp

# register as an MCP server for Claude Code
claude mcp add pentest-ai -- ptai mcp
```

## Intro

pentest-ai is a Python CLI and MCP server that lets Claude Code run verified probes, chain attack paths, and export reports for authorized testing.

- **Best for:** Authorized pentests where you want a probe-by-probe MCP loop (not a black-box scanner)
- **Works with:** Claude Code or any MCP client; `ptai` CLI for standalone/CI; common security tools auto-installed on first run (per README)
- **Setup time:** 5–15 minutes

## Practical Notes

- GitHub: 215 stars · 44 forks; pushed 2026-05-12 (verified via GitHub API).
- README introduces iterative MCP tools `list_probes` / `run_probe` / `http_request` for scoped, step-by-step driving.
- README claims 47 MCP tools and 200+ tool wrappers (nmap, nuclei, ffuf, sqlmap, gobuster, …) with auto-install on first run.

## Main

A practical “agent-safe pentest” setup looks like this:

1. **Declare scope explicitly** (domains, auth method, rate limits) in the prompt you give Claude Code.
2. Use **iterative driving**: start with `list_probes`, run one probe at a time, and only escalate when evidence supports it.
3. Prefer **proof over guesses**: keep PoCs small, reproducible, and logged; treat every finding as “untrusted” until verified.
4. Keep a clean separation between:
   - *tool execution* (ptai running probes)
   - *reasoning + coordination* (your MCP client / assistant)
5. When you need CI-style runs, switch to the CLI path (`ptai start …`) and pin provider/model settings so results are comparable run-to-run.

The big win is control: you get a repeatable probe library, plus an MCP interface that lets an LLM coordinate without inventing results.

### FAQ

**Q: Is it for authorized testing only?**
A: Yes. The README includes responsible-use warnings; only scan targets you own or have permission to test.

**Q: Do I need an API key?**
A: Not always. The README notes that when wired into Claude Code via MCP, your Claude subscription can run the engagement; otherwise you can run `ptai` with API keys (or via LiteLLM).

**Q: What should I run first?**
A: Start with low-risk recon-style probes, then iterate: run one probe, inspect evidence, and only then escalate to exploit attempts.

## Source & Thanks

> Source: https://github.com/0xSteph/pentest-ai
> License: MIT
> GitHub stars: 215 · forks: 44

---

<!-- ZH -->

## 快速使用

```bash
pip install ptai

#（可选）交互式向导，自动写入 MCP 客户端配置
ptai setup --mcp

# 注册为 Claude Code 的 MCP server
claude mcp add pentest-ai -- ptai mcp
```

## 简介

pentest-ai 是 Python CLI + MCP server，可让 Claude Code 调用 list_probes/run_probe/http_request 逐步驱动探测，输出可复现的 PoC 与报告（仅限授权测试）。

- **适合谁：** 想让 LLM“逐步驱动探测”的授权测试场景，而不是黑盒扫描器
- **可搭配：** Claude Code 或任意 MCP 客户端；也可用 `ptai` 单独跑 CI；常见安全工具首次运行会自动安装（见 README）
- **准备时间：** 5–15 分钟

## 实战建议

- GitHub：215 stars · 44 forks；最近更新 2026-05-12（GitHub API 验证）。
- README 强调可迭代驱动：`list_probes` / `run_probe` / `http_request` 让 MCP 以“逐步可审计”的方式执行。
- README 声称提供 47 个 MCP tools、200+ 安全工具封装（nmap/nuclei/ffuf/sqlmap/gobuster…），首次运行自动安装依赖。

## 主要内容

一个更“agent 安全”的渗透工作流可以这样搭：

1. **先声明范围**（域名/接口、登录方式、速率限制）再开始执行。
2. 用 **迭代驱动**：先 `list_probes`，一次只跑一个 probe，有证据再升级动作。
3. 追求 **可复现 PoC**：每个结论都以最小可复现实验证据收尾，避免“猜测式”报告。
4. 把角色分清：
   - *ptai 负责执行*（跑 probes）
   - *MCP 客户端负责推理与编排*（Claude Code / Cursor 等）
5. 需要做 CI 或定时扫描时，走 CLI 路径（`ptai start …`），并固定 provider/model，保证不同批次结果可对比。

核心收益是“可控”：你得到可重复的 probe 库 + MCP 接口，让 LLM 做编排而不是编造。

### FAQ

**只能用于授权测试吗？**
答：是的。README 有 Responsible use 提示；只对自己拥有或明确授权的目标执行。

**一定要 API Key 吗？**
答：不一定。README 提到接入 Claude Code 的 MCP 后可用订阅执行；否则可按文档配置 API key（或走 LiteLLM）。

**第一步建议怎么跑？**
答：先从低风险的探测/信息收集开始，再按证据迭代：一次只跑一个 probe，确认结果后再升级动作。

## 来源与感谢

> Source: https://github.com/0xSteph/pentest-ai
> License: MIT
> GitHub stars: 215 · forks: 44


---
Source: https://tokrepo.com/en/workflows/pentest-ai-offensive-security-mcp-for-claude-code
Author: MCP Hub